Continual Learning

Continual Learning for AI Coding Agents

Your agent forgets everything between sessions. Continual learning fixes that.

The Loop

Experience → Capture → Reflect → Persist → Apply
     ↑                                       │
     └───────────────────────────────────────┘

Quick Start

Install the hook and initialize project memory (two commands):

cp -r hooks/continual-learning .github/hooks/
mkdir -p .copilot-memory && cp hooks/continual-learning/conventions.md.template .copilot-memory/conventions.md

Initialize the SQLite DB:

sqlite3 .copilot-memory/learnings.db "
  CREATE TABLE IF NOT EXISTS learnings (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    scope TEXT NOT NULL,
    category TEXT NOT NULL,
    content TEXT NOT NULL UNIQUE,
    source TEXT,
    created_at TEXT DEFAULT (datetime('now')),
    hit_count INTEGER DEFAULT 1
  );
"

All azure-python-* agents load .copilot-memory/conventions.md and the DB automatically at the start of every session.

Two-Tier Memory

Global (~/.copilot/learnings.db) — follows you across all projects:

• Tool patterns (which tools fail, which work)
• Cross-project conventions
• General coding preferences

Local (.copilot-memory/learnings.db) — stays with this repo:

• Project-specific conventions
• Common mistakes for this codebase
• Team preferences

How Learnings Get Stored

Automatic (via hooks)

The hook observes tool outcomes and detects failure patterns:

Session 1: bash tool fails 4 times → learning stored: "bash frequently fails"
Session 2: hook surfaces that learning at start → agent adjusts approach

Agent-native (via store_memory / SQL)

The agent can write learnings directly:

INSERT INTO learnings (scope, category, content, source)
VALUES ('local', 'convention', 'This project uses Result<T> not exceptions', 'user_correction');

Categories: pattern, mistake, preference, tool_insight

Manual (memory files)

For human-readable, version-controlled knowledge:

# .copilot-memory/conventions.md
- Use DefaultAzureCredential for all Azure auth
- Parameter is semantic_configuration_name=, not semantic_configuration=

Compaction

Learnings decay over time:

• Entries older than 60 days with low hit count are pruned
• High-value learnings (frequently referenced) persist indefinitely
• Tool logs are pruned after 7 days

This prevents unbounded growth while preserving what matters.

Best Practices

1. One step to install — if it takes more than cp -r, it won't get adopted
2. Scope correctly — global for tool patterns, local for project conventions
3. Be specific — "Use semantic_configuration_name=" beats "use the right parameter"
4. Let it compound — small improvements per session create exponential gains over weeks